Use OpenAI models through one API, with pricing in Sri Lankan rupees and usage on one monthly invoice. No separate OpenAI account or foreign card needed.
| Model | Input / 1M | Output / 1M | Context | Best for | Processed in | |
|---|---|---|---|---|---|---|
| LKR 3,330 | LKR 16,650 | 1.05M | Hard maths, science and agent tasks | Global | ||
| Cached input | LKR 333LKR 4,163 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 1,665 | LKR 8,325 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 6,660 | LKR 24,975 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6 Astra | ||||||
| LKR 2,664 | LKR 13,320 | 1.05M | Hard science and maths | Global | ||
| Cached input | LKR 133LKR 3,330 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 1,332 | LKR 6,660 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 5,328 | LKR 19,980 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6.2 Astra | ||||||
| LKR 1,332 | LKR 6,660 | 1.05M | Long coding tasks | Global | ||
| Cached input | LKR 133LKR 1,665 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 666 | LKR 3,330 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 2,664 | LKR 9,990 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-5.6 Sol | ||||||
| LKR 666 | LKR 3,330 | 1.05M | Everyday work and agents | Global | ||
| Cached input | LKR 33.30LKR 833 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 333 | LKR 1,665 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 1,332 | LKR 4,995 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6.2 Sol | ||||||
| LKR 666 | LKR 3,330 | 1.05M | Production coding agents | Global | ||
| Cached input | LKR 33.30LKR 833 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 333 | LKR 1,665 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 1,332 | LKR 4,995 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6.1 Sol | ||||||
| LKR 666 | LKR 3,330 | 1.05M | Products already built on GPT-6 Sol | Global | ||
| Cached input | LKR 66.60LKR 833 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 333 | LKR 1,665 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 1,332 | LKR 4,995 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6 Sol | ||||||
| LKR 583 | LKR 4,662 | 400K | Terminal coding agents | Global | ||
| Cached input | LKR 58.28 | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Everything about GPT-5.3-Codex | ||||||
| LKR 66.60 | LKR 400 | 1.05M | High-volume extraction | Global | ||
| Cached input | LKR 6.66LKR 83.25 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 33.30 | LKR 200 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 133 | LKR 599 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-5.6 Luna | ||||||
| LKR 66.60 | LKR 416 | 400K | Classification and tagging at scale | Global | ||
| Cached input | LKR 6.66 | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 33.30 | LKR 208 | Results within 24 hours, for work nobody is waiting on. | |||
| Everything about GPT-5.4 nano | ||||||
| LKR 64.94 | LKR 260 | 131K | Maths, science and logic | Global | ||
| Cached input | LKR 32.47 | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Everything about gpt-oss-120b | ||||||
| LKR 33.30 | LKR 167 | 1.05M | High-volume extraction and routing | Global | ||
| Cached input | LKR 3.33LKR 41.63 to write | — | Repeat a long prompt prefix and those tokens bill at this rate. Automatic. | |||
| Batch | LKR 16.65 | LKR 83.25 | Results within 24 hours, for work nobody is waiting on. | |||
| Long requests | LKR 66.60 | LKR 250 | Above 272,000 input tokens, the whole request is billed at these rates. | |||
| Everything about GPT-6 Luna | ||||||
| LKR 30.30 | LKR 108 | 131K | Quick reasoning steps | Global | ||
| Standard rates only. Everything about gpt-oss-20b | ||||||
| LKR 4,995 characters | — | — | Real-time voice replies | Global | ||
| Standard rates only. Everything about TTS-1 | ||||||
| LKR 4,810 audio seconds | — | — | Transcripts and captions | Global | ||
| Standard rates only. Everything about Whisper large-v3-turbo | ||||||
| LKR 43.29 | — | 8K | Multilingual search and retrieval | Global | ||
| Batch | LKR 21.65 | Free | Results within 24 hours, for work nobody is waiting on. | |||
| Everything about text-embedding-3-large | ||||||
| LKR 6.66 | — | 8K | Low-cost search and retrieval | Global | ||
| Batch | LKR 3.33 | Free | Results within 24 hours, for work nobody is waiting on. | |||
| Everything about text-embedding-3-small | ||||||
Rupee prices are our published US dollar rates converted at LKR 333 to the dollar, our rate for November 2026 — the same rate your invoice uses.
The newest everyday model, for writing, analysis and agents at volume.
The flagship, for advanced science and maths, where the answer can wait.
Production coding agents, professional documents and spreadsheet work.
The cheapest GPT-6, for sorting, extracting and routing at volume with a million-token context.
Classifying, tagging and ranking large volumes of text.
Search and retrieval over your own documents, including languages other than English.
Transcripts and captions in 99 languages, Sinhala and Tamil among them.
The Roar AI gateway speaks the OpenAI API format. Point the OpenAI SDK at Roar AI and name a OpenAI model; moving to another provider’s model later is a change to the model name.
from openai import OpenAI
client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")
reply = client.chat.completions.create(
model="gpt-6.2-sol",
messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)OpenAI models are processed on international infrastructure, outside Sri Lanka. None of them is hosted locally.
If a workload’s data has to stay in the country, Qwen 3.8 27B LK runs on our own server in Colombo.
Yes. Every OpenAI model listed on this page is available from Sri Lanka through Roar AI’s API, on the same key as every other model on the platform.
Yes. OpenAI usage is billed to one monthly invoice in LKR, together with any other models your team uses.
No. One Roar AI key reaches every OpenAI model listed here, so there is no separate OpenAI account or foreign card to set up.
No. OpenAI models are processed outside Sri Lanka. Roar AI’s locally hosted model, Qwen 3.8 27B LK, is a separate model and is labelled as such.
One key for every OpenAI model and every other model, on one monthly invoice in LKR.